DNB Working Paper
No. 222 / September 2009
Gabriele Galati, Steven Poelhekke and Chen Zhou
DNB W o r k i n g P a p e r
Did the crisis affect inf lation
expectations?
Did the crisis affect inflation expectations?
Gabriele Galati, Steven Poelhekke and Chen Zhou *
* Views expressed are those of the authors and do not necessarily reflect official
positions of De Nederlandsche Bank.
Working Paper No. 222/2009
September 2009
De Nederlandsche Bank NV
P.O. Box 98
1000 AB AMSTERDAM
The Netherlands
Did the crisis affect inflation expectations?
by Gabriele Galati, Steven Poelhekke and Chen Zhou ∗
DNB, August 2009
Abstract
We investigate whether the anchoring properties of long-run inflation expectations in the United
States, the euro area and the United Kingdom have changed around the economic crisis that erupted in
mid-2007. We document that in these three economies, expectations measures extracted from
inflation-indexed bonds and inflation swaps became much more volatile in 2007. Moreover, their
sensitivity to news about inflation and other domestic macroeconomic variables – a measure of
anchoring – increased first during the oil price rally in 2006–07, and then during the heightened
turmoil triggered by the collapse of Lehman Brothers. Liquidity premia and technical factors have
significantly influenced the behaviour of inflation-indexed markets since the outburst of the crisis. We
show, however, that these factors did not contaminate the relationship between macroeconomic news
and financial market-based inflation expectations at the daily frequency. By testing for structural
breaks we conclude that in the United States, the euro area and the United Kingdom, long-run inflation
expectations have become less firmly anchored during the crisis.
JEL Classification: E31, E44, E52, E58.
Key words: monetary policy, inflation and inflation compensation, anchors for expectations, crisis,
liquidity.
∗ Research Department, De Nederlandsche Bank. Views expressed are those of the authors and do not necessarily reflect
official positions of De Nederlandsche Bank nor of the Eurosystem of central banks. We would like to thank Maria
Demertzis, Refet Gürkaynak, Peter Hördahl, Pierre Lafourcade, Martijn Schrijvers, Nikola Tarashev and participants at
seminars at DNB, Vrije Universiteit Amsterdam, the 2009 INFINITI conference and the 28th SUERF Colloquium for useful
discussions and comments. Any remaining errors are our own.
1. Introduction
“Ultimately, the firm anchoring of inflation expectations remains the best way to check the
appropriateness of monetary policy in an uncertain environment.” (Bini-Smaghi, 2009)
After rising sharply in 2007 and the first half of 2008 – against the background of rallying commodity
and food prices – global inflation went on a marked downward trend when the macroeconomic
consequences of the financial crisis became visible in the fall of 2008. A few months later, price
changes turned negative in a number of economies, sparking an intense debate on whether deflationary
risks were likely to materialise and what steps would be needed to avoid deflation (Jeanne, 2009). The
discussion centred to an important extent on how inflation expectations have been affected by the
crisis, and in particular on whether the anchoring properties of long-run inflation expectations changed
as the crisis unfolded (Svensson, 2009). Our paper attempts to answer this question, by examining the
behaviour of long-run inflation expectations in three major economies – the United States, the euro
area and the United Kingdom – in the years around the crisis. This is the first empirical study of the
behaviour of expectations during a crisis.
We proceed in two steps. We first examine the time series behaviour of two types of measures of longrun inflation expectations – survey-based measures and measures extracted from financial market
instruments – for the United States, the euro area and the United Kingdom. We argue that the semiannual frequency of the former make these measures less suited for an analysis of changes that might
have occurred within a relatively short time span. By contrast, information extracted from financial
markets is available at daily or even intra-day frequency, and therefore useful for our empirical
exercise. Markets for inflation-indexed bonds and swaps in these economies are the most liquid in
“normal” times and have enjoyed sufficient liquidity to allow extracting “reasonable” measures of
inflation compensation from them.
To verify whether inflation expectations are firmly anchored, we first need to check whether they vary
around central banks’ inflation objectives. In addition, following the approach of Gürkaynak et al.
(2005), Gürkaynak et al. (forthcoming) and Beechey et al. (2007), it is necessary to investigate the
reaction of financial market-based measures to news about inflation and other domestic
macroeconomic variables. The idea is that if long-run inflation expectations are perfectly anchored,
they should not react to the arrival of news but rather be stable around the central banks target for
inflation. In particular, we test whether the reaction of expectations to news has changed since the
outburst of the crisis. Statistically and economically significant evidence of such a change would
indicate that the crisis has influenced market participants’ perception of the Fed, the Bank of England
and the ECB’s commitment to price stability. One conjecture is that the unprecedented monetary
2
easing (through both conventional and non-standard monetary policies), coupled with the
accumulation of a huge fiscal debt, may have undermined market participants’ confidence in the
ability of central banks to keep inflation at target in the longer-run.
We find several interesting empirical results. First, survey-based measures of long-run inflation
expectations in the three economies remained fairly stable within the central bank’s comfort zone both
during the oil price rally in 2006–07 and as the crisis unfolded. We argue that the frequency at which
inflation surveys are conducted is too low to allow inferences on the stability of anchoring properties
during a relatively short time span.
Second, using financial market-based measures we find evidence that anchoring properties have
changed, starting before the crisis. These measures have become much more volatile around mid2007. Moreover, structural break tests show that inflation expectations have become more sensitive to
macroeconomic news, particularly during the heightened turmoil triggered by the collapse of Lehman
Brothers.
Third, while liquidity premia and technical factors appear to have significantly influenced the
behaviour of inflation expectations derived from inflation-indexed bonds or inflation swaps since the
outburst of the crisis, they do not appear to have contaminated the relationship at the daily frequency
between macroeconomic news and financial market-based inflation expectations. We therefore feel
confident that financial market-based measures of long-run inflation expectations allow us to draw
inferences about their anchoring properties during the crisis.
The remainder of the paper is organised as follows. Section 2 provides brief overview of the relevant
literature on the anchoring of inflation expectations. Section 3 describes two measures of long-run
inflation expectations: survey-based measures and measures backed out from financial instruments.
Section 4 discusses the role of liquidity and technical factors in financial market measures of inflation
expectations. Section 5 presents our empirical model and the main results. Section 6 concludes.
2. Literature review
While expectations and credibility play a central role in the theoretical literature, there is little
theoretical work on the concept of “anchoring” of inflation expectations. In standard macroeconomic
theory, if the central bank’s objective function is known and constant, the rational expectations
hypothesis implies that long-run inflation expectations do not change over time in response to the
arrival of new information. In fact, Del Negro and Eusepi (2009) show that standard medium scale
DSGE models have difficulties explaining the evolution of inflation expectations, and that the fit is
even worse when the assumption of perfect information is relaxed. In recent years, a series of papers
3
departed from the rational expectations hypothesis and the assumption of a known and constant central
bank objective (e.g. Orphanides and Williams, 2005; Brazier et al., 2008; Demertzis et al., 2007,
2008). This approach allows more realistic models of the link between inflation expectations and
underlying inflation.
A key element of these models is the relationship between inflation expectations and shocks to the
economy. The higher the sensitivity of expectations to these shocks, the more successful monetary
policy will be. In Orphanides and Williams (2005) agents do not know the true model of the economy
but rather constantly update their estimates based on all information available to them. As a result,
inflation expectations are sensitive to economic shocks. Orphanides and Williams (2005) introduced
central bank communication in their model and find that with learning, successful communication
reduces the sensitivity of inflation expectations to actual inflation.
A similar idea is found in Demertzis et al. (2007), who modelled monetary policy as an information
game, in which individual agents have to interpret new (publicly available and private) information
when they form ex ante expectations about future long-term inflation. In their model, ex post inflation
is a function of the monetary policy chosen by the central bank to pursue its objectives and the average
of all individual expectations. It is then optimal for agents to form expectations based on three
elements: monetary authorities’ objectives and their policy decisions; shocks that occur after these
decisions; and the average of individual inflation expectations. Once the central bank communicates
its inflation objective to the public — such as the ECB’s operational definition of price stability as
below but close to 2% – agents can either form their expectations based on the above three elements
or, alternatively, coordinate their expectations on that target. They derived a time-varying parameter
that captures the credibility of the central bank’s target. If the target’s credibility is sufficiently high,
individual agents will focus their expectations on that target.
In a companion paper, Demertzis et al. (2008) took their model to the data and estimate the degree to
which long-run inflation expectations in the United States have been anchored to the Fed’s objective.
They tested whether long-run inflation expectations– derived from the Fed’s FRB model or quarterly
survey-based measures – are influenced by short-run inflation dynamics. They estimated a timevarying parameter (λ) that measures the extent to which inflation expectations are anchored across
quarters. They found that in recent years, the anchorness of expectations in the United States has
weakened but only slightly, without compromising the Fed’s credibility.
Agents may also use rules of thumb (“heuristics”) to make inflation forecasts. Brazier et al. (2008)
consider two heuristics: one is based on lagged inflation and the other on an inflation target announced
4
by the central bank. In their model, agents switch between these two heuristics based on an imperfect
assessment of how each has performed in the past.
The empirical literature on drivers of inflation expectations – surveyed carefully in a paper by Clark
and Davig (2008) – has highlighted the role of macroeconomic variables. 1 The periodical
announcements on the state of the economy and forecasts released by various (statistical) offices and
agencies form a steady source of new information. To the extent that the new information is
unanticipated, beliefs about future inflation may be updated. If expectations are perfectly anchored –
i.e. the central bank credibly commits to its inflation objective – long-run inflation expectations should
not be responsive to news about actual inflation, or more general about macroeconomic conditions.
A number of recent studies documented the anchoring of long-run inflation expectations in a number
of countries. One strand of the literature relies on inflation surveys. Levin et al. (2004) analysed the
behaviour of private-sector inflation forecasts at horizons up to ten years – measured by quarterly
Consensus forecasts – in United States and the euro area over the period 1994–2003. They found that
expectations were highly correlated with a three-year moving average of lagged inflation. By contrast,
in industrial countries that have adopted inflation targeting (United Kingdom, Sweden, Canada,
Australia and New Zealand), inflation expectations were found not to be sensitive to actual inflation.
Levin et al. (2004) concluded that inflation targeting has played a significant role in anchoring longrun inflation expectations. Paloviita and Viren (2005) found that inflation expectations, proxied by
OECD inflation forecasts, respond to changes in output and actual inflation. The results are based on a
simple VAR model with inflation, inflation expectations, estimated with pooled annual data for euro
area countries over the period 1979–2003. Clark and Nakata (2008) showed that in the United States,
inflation expectations appeared to be slightly better anchored in recent years compared to 20 or more
years ago. In particular, they found evidence of a declining impact of unexpected increases in inflation
on long-term expectations.
A second strand of the literature extracted inflation expectations from inflation-indexed financial
market instruments, and looked at the relationship between inflation expectations and macroeconomic
variables at high (daily or intraday) frequency (Swanson, 2006). Gürkaynak et al. (2003, 2005) derived
inflation expectations from inflation-indexed bonds and examined their sensitivity to surprises about
macroeconomic announcements at the daily frequency. To test for anchoring of inflation expectations,
they regressed daily inflation expectations on a set of macroeconomic news variables. They found that
between 1990 and 2002, long-run inflation expectations in the United States were not perfectly
anchored. This analysis was extended by Gürkaynak et al. (forthcoming) and Gürkaynak et al. (2006),
1 Another strand of literature relies on economic experiments (e.g. Marimon and Sunder, 1994; Hommes et al, 2005, 2007;
Adam, 2007).
5
who documented that long-run inflation expectations are more solidly anchored in the United
Kingdom, Sweden and Canada – countries that adopted formal inflation targets. Consistently with
Levin et al. (2004), they concluded that a numerical inflation target has helped anchoring long-term
inflation expectations.
Beechey et al. (2007) followed the same methodology to compare the anchoring properties of long-run
inflation expectations in the United States and the euro area over the period 1 June 2003 – 31
December 2006. They found that surprises about monetary policy decisions and macroeconomic data
releases – the core CPI but also indicators of economic activity such as the National Association of
Purchasing Managers (NAPM) index or non-farm payrolls – have significant effects on US forward
inflation compensation at different horizons. By contrast, long-term inflation compensation does not
significantly react to any news about price or output developments in the euro area. Beechey et al.
(2007) concluded that long-run inflation expectations are more firmly anchored in the euro area than in
the United States.
This paper builds on the literature by developing a method to assess possible changes in anchoring of
inflation expectations around the recent crisis. 2 Mounting inflationary pressure due to booming
commodity prices in the run-up to the crisis might have caused inflation expectations to drift.
Similarly, the crisis may well have led to drifting inflation expectations in the wake of unsurpassed
monetary and fiscal expansion.
3. Measuring inflation expectations
Survey based measures
There are two main approaches to measuring long-run inflation expectations. The first approach relies
on inflation surveys. 3 A frequently used data source, Consensus Economics, provides semi-annual
data on expectations of a panel of some 30 professional forecasters six to ten years ahead, for a
number of countries. There are also survey data for somewhat shorter horizons. For example, at a
horizon of 5 years, the ECB Survey of Professional Forecasters (SPF) collects, on quarterly basis,
forecasts by a panel of some 70 professional forecasters on euro area HICP.
2 Using a similar empirical approach, Galati, Poelhekke and Zhou (2008) tested whether the sharp increase in HICP inflation
in the year before the crisis, which was underpinned by a sharp rally in commodity and food prices, led to an un-anchoring of
long-run expectations in the euro area. They backed out long-run expectations inflation expectations from euro denominated
nominal and inflation-indexed swaps, and investigated the reaction of these measures to news about inflation and other
macroeconomic variables in the main euro area countries. They found some evidence that long-run inflation expectations
started to drift away from the ECB's anchor in the course of 2007.
3 A detailed analysis of the properties of these measures is provided in ECB (2006). For a careful analysis of the properties of
survey measures, see Clark and Davig (2008).
6
Figure 1 shows the time series of SPF data on expectations of euro area inflation over a 5-year
horizon. The graph highlights that neither the oil price rally in the years preceding the crisis nor the
crisis itself have had a visible influence on inflation expectations at the 5-year horizon.
However, survey measures have several important shortcomings. First, given their low frequency,
survey measures appear well-suited for analysing longer-run properties of inflation expectations but
not for identifying the existence and timing of breaks in the expectation formation over a short
horizon. Second, survey results may not be reliable to the extent that respondents do not have to act on
the basis of their responses – i.e. “do not put their money where their mouth is”. 4 Third, as shown by
Van der Klaauw et al. (2008), survey results are sensitive to the wording of the questions. Fourth,
different types of survey measures may produce very different results. Mankiw et al. (2003), for
example, looked at 50 years of data on inflation expectations in the United States, and documented
substantial disagreement among both consumers and professional economists about expected future
inflation. They found that this disagreement varied substantially through time, depending on the level
of inflation, the absolute value of the change in inflation, and relative price variability.
Financial market based measures
A second approach to measuring long-run expectations consists in using financial market-based
measures. In a number of countries, bonds or interest rate swaps that are linked to some measure of
domestic inflation are actively traded (Deacon et al., 2004, and JP Morgan, 2008). These instruments
can be combined with nominal bonds or nominal interest rate swaps to back out financial markets’
inflation expectations. The main advantage of this type of measure is that, given its high frequency, it
allows examining changes in the behaviour of expectations over a relatively short horizon. It is
therefore most useful in investigating whether the financial crisis has affected the anchoring of
inflation expectations to the Fed’s, the ECB’s, and the Bank of England’s inflation objectives.
In this paper we derived inflation expectations from nominal and inflation-indexed financial
instruments. In particular, we considered inflation-indexed bonds for the United States (Treasury
Inflation-Protected Securities, TIPS) and the United Kingdom (inflation-linked Gilts), and inflation
swaps for the euro area. These instruments are actively traded and the most liquid ones among
inflation-indexed products. 5 For example, market commentary indicated that the monthly trading
volume of 10-year euro area inflation swaps averaged around 6bn in 2007 (JP Morgan, 2008). Buyers
of inflation swaps primarily include insurance companies and pension funds, which suffer a loss of
4 This point is emphasised in the experimental economics literature on inflation expectations (Smith, 1982).
5 Alternatively, we could have used nominal and inflation-indexed bonds also for the euro area, which in recent years have
become increasingly liquid. Our preference for swaps is motivated by their greater liquidity along the whole maturity
spectrum, particularly in the earlier years in our sample period. Beechey et al (2007) note that inflation compensation derived
from swaps and TIPS are very similar. For a more extensive discussion of the difference between inflation-indexed bonds
and inflation swaps, see Deacon et al. (2004) and JP Morgan (2008).
7
income if the actual inflation rate increases. Typical inflation swap sellers include firms whose income
is linked to inflation while their expenses are not, or only to a lesser degree, such as public authorities,
utilities, real estate companies, or distribution companies. By selling inflation in an inflation-linked
swap, they are able to protect future income linked to inflation. Data for the yield curves of UK
inflation-indexed bonds were taken directly from the Bank of England, while data on US TIPS and
euro area swaps were obtained from Bloomberg. 6
Here we describe in detail the method followed to extract inflation expectations from swaps markets
but the main features apply also to the approach we followed for bond markets. Inflation expectations
can be measured by the difference between one-year zero-coupon forward rates of inflation swaps and
nominal interest rate swaps. Since we focus on long-term inflation expectations, we chose one-year
zero-coupon forward rates ending ten year ahead. Hence, at day t inflation expectation are measured
by
(1)
is
in
f t = f t,10
− f t,10
,
is
where f t,10
and f t in,10 denote the one-year zero-coupon forward rates ending ten year ahead for
inflation swaps and interest swaps respectively.
We collected daily data on euro area swap markets for the period 23 June 2004 to 24 March 2009. For
each day, swap rates are available for different maturities, allowing us to estimate a whole yield curve.
The forward rates ending ten year ahead can be calculated from the swap rates with 9- and 10-year
maturities as
(1+ y10 )
f t,10 =
9
(1+ y 9 )
10
(2)
−1,
where y 9 and y10 are the 9-year and 10-year swap rates, respectively. 7
One concern is that since the market for swaps with a 9-year maturity may be less liquid than the
market for 10-year maturity, estimates of forward rates based on 9-year swaps may exhibit a high level
of noise. In order to filter out this noise, we followed a standard technique of the finance literature –
the Nelson-Siegel method (see, e.g. Nelson and Siegel, 1987) – to estimate a smooth yield curve for
each day. Details on this method are presented in Appendix 1. Söderlind and Svensson (1997) argued
that estimating yield curves by a simple curve fitting, as we did, rather than by a structural model for
6 See http://www.bankofengland.co.uk/statistics/yieldcurve/index.htm.
7 The same formula to calculate the forward rates applies to both inflation swaps and interest swaps. This argument holds for
the rest of the section.
8
interest rate dynamics is appropriate when the purpose is to extract market expectations about future
interest rates without making additional assumptions about the model structure.
We obtained two smoothed yield curves for each day: one for inflation swaps, the other one for
interest rate swaps. From these, we got the smoothed 9- and 10-year swap rates. We then estimated 10year ahead forward rates, from which we derived the inflation expectation measures as a difference
between inflation swaps and nominal interest rate swaps. Note that our smoothing procedure is applied
for each specific day but it gives also a smoother measure of inflation expectations across time. This
smoothing effect can be explained in terms of a smaller impact of liquidity effects. Throughout the
paper, we used the smoothed series as a measure of long-run inflation expectations.
Figure 1 shows that our measure of long-run inflation expectations for the euro area based on
inflation-indexed swaps differs significantly from the inflation expectations at the 5-year horizon
given by the SPF. While the SPF measure were very stable and close to the ECB’s objective of 2%
medium-term inflation over the whole sample period, inflation expectations from swap markets swung
between 2.5% and 1% and became much more volatile after the onset of the crisis. Similarly, financial
market and survey based measures of long-run inflation expectations also differ visibly for the United
States and the United Kingdom.
4. The role of liquidity premia and technical factors
One potential reason for the visible difference between survey and financial market based measures of
inflation expectations is that the latter may be contaminated by other factors, especially during the
crisis. 8 Break-even rates, i.e. the difference between nominal and inflation-indexed bonds, can be
decomposed into four main factors: expected inflation, inflation risk premia, liquidity premia, and
technical factors (Hördahl, 2009). According to Hördahl (2009), the same applies to a much lesser
extent to the difference between nominal and inflation-indexed swaps. One example of these technical
factors are sudden portfolio shifts by leveraged investors that may affect nominal and inflationindexed bond markets but are unrelated to changing views about future economic fundamentals. One
critical assumption of exercises that back out measures of inflation expectations from financial
instruments is that the last two factors do not respond to macroeconomic news at the daily frequency.
Changes in far-horizon break-even rates can then be interpreted as a revision of market participants’
long-run inflation expectations or inflation risk in response to new information, indicating that
inflation expectations are not solidly anchored. 9
8 See e.g. Barclays Capital Research (2008).
9 The inflation risk reflects both the volatility of inflation expectations as well as market participants’ attitude towards risk. If
inflation expectations are firmly anchored, none of these two components should react to new information.
9
In “normal” times this assumption appears plausible. Dudley et al. (2009) documented the deepening
of TIPS markets by looking at trading volumes, bid-ask spreads and estimates of illiquidity premia.
Beechey et al. (2007) documented that the announcement effects on inflation expectations measured
using TIPS or inflation swaps persist for about one business week. They interpreted this as evidence
that the reaction of the dependent variable is not driven primarily by liquidity effects. Beechey et al.
(2008) decomposed US nominal yields into three components – nominal yields, real yields, and the
spread between these two (i.e. inflation compensation) – and then estimated the effect of news on
these three components using intra-day data. They found that different types of news – about prices,
the real economy or monetary policy – have quite different effects on real rates and rates of inflation
compensation. In particular, only news about prices affect inflation compensation. They also tested
whether the impact of news has changed over time, since the market for inflation-indexed bonds in the
United States – Treasury Inflation Protected Securities (TIPS) – has deepened. Their evidence
suggests that the reaction of long-term inflation compensation to inflation news has not changed
between 17 February 2004 and 13 June 2008, implying no significant change in inflation expectations’
anchoring properties during this period. 10
During the current crisis, the assumption that changes in market liquidity and technical factors may
contaminate the behaviour of financial market-based measures of inflation expectations appears much
less innocuous. Financial markets, including bond and swap markets, experienced pronounced swings
in volatility and liquidity. In these circumstances, there was evidence that yields on nominal and
inflation-indexed bonds (and swaps) were driven not just by expectations about future inflation but
also by high and volatile liquidity premia and technical factors related e.g. to hedging activity (Fender
et al., 2009; Hördahl, 2009).
Note that by construction, our measure of inflation compensation filtered out part of the noise. In
particular, by taking the difference between nominal and inflation-indexed bonds (or swaps), we
purged the effect of liquidity and technical factors that affect both markets in a similar way. For
example, if on a particular day there is a sudden broad portfolio shift out of fixed income markets in
general and into equity markets, nominal and inflation-indexed bond yields could both be affected in a
similar fashion. Moreover, we focus on day-to-day changes in inflation compensation, and the relative
liquidity of the nominal and inflation-indexed markets may not change substantially from day to day.
This discussion suggests that simple graphical evidence on the behaviour of break-even rates can be
misleading: the “true” inflation expectations may be anchored even though financial market based
10 By contrast, using daily data from early 1999 to early 2004, Beechey et al. (2008) compared the performance of their
model before and after 2004 and found evidence that inflation compensation became less sensitive to news after 2004. They
interpreted this result as suggesting that the improved liquidity and functioning of the TIPS market since 2004 may have
allowed TIPS yields to become more responsive to new information.
10
measures are not close to the central bank’s objective if the influence of liquidity and technical factors
is not filtered out properly. In addition, even if our expectation measure is “accidentally” close to the
central bank’s objective, it is not sufficient to conclude that inflation expectations are firmly anchored.
In this case, the movements of expectations are influenced by macroeconomic (and other) news, which
happen to drive inflation expectations close to the central bank’s objective. To get more accurate
evidence on the anchoring of long-run inflation expectations, we therefore turn to examine their
response to macroeconomic news.
5. Empirical results on breaks and anchoring
In order to test whether inflation expectations became unanchored during the crisis, we examined the
impact of news on HICP inflation and other macroeconomic variables on our measures of inflation
expectations. Our focus is on testing for structural breaks while improving estimation by dealing
explicitly with liquidity effects. Our sample period is 23 June 2004 – 23 March 2009, and includes
almost two crisis years during which liquidity effects may have been especially important.
Following the approach developed by Gürkaynak et al. (2006) and Beechey et al. (2007) – who
compared the credibility of the Fed, the ECB and inflation targeting central banks – we captured news
by the difference between actual releases of the main euro area macroeconomic variables and values
anticipated by market participants according to surveys conducted by Bloomberg and JP Morgan. This
is a common method in the literature, although Rigobon and Sack (2008) found that it tends to
underestimate the responses to true news because of measurement errors. In fact, Bartolini et al.
(2008) discussed shortcomings of survey-based measures of news but conclude that this approach may
be the only one available in practice.
We expect that data releases on inflation variables are most important. However, without empirical
guidance from the literature, we have few other priors on what type of information influences inflation
expectations. Other macroeconomic variables – such as GDP growth, business confidence indicators,
the unemployment rate or wage growth – may also give indications about possible inflationary
pressures. The used the same macro-announcements as in Beechey et al. (2007). 11 For the euro area,
we concentrated on data releases for the three main economies – Germany, France, and Italy – since
these are most likely to have a primary influence on views on future euro area inflation. Considerably
more news variables are available through Bloomberg and JP Morgan for the United Kingdom and the
United States. Appendix 2 lists all available variables.
11 One difference with Beechey et al (2007) is that we do not include US news in the regressions where the dependent
variable is euro area or UK inflation expectations.
11
We regressed our measure of long-run inflation expectations on a constant, a set of macroeconomic
news variables and a set of control variables, according to the following model:
(3)
Δf t = α + β Xt + γ Zt + ε t
where the dependent variable Δf t = f t − ft −1 is the change, from closing of the markets at day t-1 to
closing on day t, in one-year inflation compensation ten years ahead. The explanatory variables Xt are
a vector of news variables on various measures of the state of the economy. Most macroeconomic
news arrives at 8:30 am, before stock markets open. 12 Our expectations variable therefore measures
the change in inflation expectations between the end-of-day swap quote of the day before macro news
arrive (t-1), and the end-of-day quote on the same day on which the news arrive (t).
Zt is a vector of control variables intended to capture the influence that shorter-term changes in
liquidity premia and technical factors unrelated to inflation expectations may have on swap rates. In
particular, our control variables are useful in purging the effect of liquidity premia and technical
factors that are related to shocks that broadly hit financial markets. For example, a sudden increase in
financial stress on a particular day due to news about the collapse of a major financial player may
induce a broad flight to liquidity. This could benefit nominal bonds at the expense of other assets such
as inflation-index bonds. Our preferred control variable is the implied volatility of bond yields but we
checked that our results were robust to using four alternative variables that measure market liquidity:
the Chicago Board Options Exchange Volatility Index (VIX), a widely used measure of the implied
volatility of S&P 500 index options; the euro bund implied volatility; the on-the-run off-the-run spread
(a commonly measure of bond market liquidity); and analogously for the euro area, the KfU-bund
spread. 13 All these non-news controls were first differenced as well, because a change in implied
volatility should lead to an additional change in our measure of inflation expectations through a
liquidity premium. We also controlled for day-of-the-week effects but these turned out not to be
statistically significant. 14
We interpret a high R2 – to the extent that it is driven by significant coefficients on the variables Xt –
as evidence that expectations are weakly anchored. A low R2 conversely implies well anchored
inflation expectations. A change in explanatory power of the model during the sample period would
then indicate that anchoring properties of inflation expectations have changed. In particular, we
verified whether the sensitivity of inflation expectations to news about inflation and other
12 See for example JP Morgan (2009).
13 A description of this variable is given in Hördahl (2009).
14 For reasons of space, the results for these dummies are not reported here. All variables used in the regressions are
stationary.
12
macroeconomic variables has increased since 2006, when first commodity and food prices started to
rally and then the financial crisis erupted.
To detect such changes, we need to test for a structural break. We used the Chow test, which verifies
whether the structure of the model changes on a certain day. However, it is an ex post test in the sense
that the timing of the break must be chosen with some prior knowledge. Chow tests for dates that are
close to the real structural break will also give statistically significant results. To get a more precise
indication of the timing of the structural break, we therefore used a rolling version of the Chow test.
Changing the date at which we split the sample gives us a measure of when, if at all, the model starts
to predict changes in inflation expectations. If we find that the model fit increases after the break we
may conclude that expectations have become less anchored from that point on. Note that we adjusted
the Chow test statistic so that we capture only the change in model performance due to news
variables. 15
As a further step, we considered the most conservative dating of a structural break, given by the
maximum of the test statistics from the rolling Chow tests (see Zeileis et al., 2003). We examined the
change over time in both the F-test value and p-value to assess the timing of a structural break. 16
Our results – summarised separately for the United States, the euro area and the United Kingdom in
Figures 3, 4 and 5 – indicate that the sensitivity of inflation expectations to news has indeed changed
during the sample period. Each figure plots on the left axis the p-value of the Chow test for each day,
and on the right axis the actual value of the test statistic. Several results stand out. First of all, the test
statistics (and, inversely, the p-values) all increase towards end-2008 and early 2009, around the time
when the financial crisis peaked. The closer we get to the period of heightened financial stress – as
highlighted by the dotted vertical line, the date at which Lehman Brothers filed bankruptcy – the more
different the model performs in the later sample period compared to the first period. This indicates that
in all three areas inflation expectations have changed sensitivity over time to macroeconomic news.
Secondly, we found a break in each economic area. The dates at which the Chow test becomes
significant for the first time at the 99% confidence level are highlighted in the graphs by vertical
dashed lines, and are reported in Table 1. We identified a first statistically significant change in the
behaviour of inflation expectations around December 2006 for the euro area and March 2006 for the
United States. In the United Kingdom we identified a break much later, in April 2008.
15 If liquidity premia are important, then the implied volatility variables may also cause the Chow test to flag a break, just
because volatility increased during the crisis period. To prevent this, we replaced the sum of squared residuals from the three
regressions (entire period, before and after potential break point) in the test statistic with one minus their partial R-squared –
where the explanatory effect of the control variables was partialled out – multiplied by their respective total sum of squares.
16 We truncate the sample by 100 observations on each side to allow for feasible testing near the beginning and end of the
sample.
13
Thirdly, the test statistics show a hump shaped pattern around the crisis period in all three economies,
suggesting strong evidence for a break at the height of the crisis and decreasing evidence in the
following months.
Table 1 shows, in addition to the timing of the first break identified by Chow tests, dates on which the
Chow test statistics is maximized. Using this more conservative method of dating structural breaks in
the relationship between inflation expectations and news we find more similarities across the United
States, the euro area and the United Kingdom. For all three economies, the maximum Chow test
statistic can be found between September and November of 2008, a period that is commonly
considered as the height of the financial crisis, and encompasses the collapse of Lehman Brothers on
15 September 2008.
As a robustness check, we used the method developed by Andrews (1993), which tests whether a
structural break exists within a short period without knowing the break point. 17 The essence of these
tests is to look at the test statistics yielded by Chow-tests on each day within the period under
consideration, and construct new unified testing statistics from them. This helps to narrow down the
location of the structural break. Once we identified the day on which the F-test statistic of our rolling
Chow tests reaches their maximum, we applied the Andrews test to verify whether it identifies
significant structural breaks around those days. The Andrews test indeed confirmed that at a 99%
confidence level, a structural break exists on the days on which the Chow test statistics are maximized
for all three economic areas.
Having identified structural breaks in the relationship between inflation expectations and inflation and
macroeconomic news, we looked at the fit of the model in each sub-sample to tell us if inflation
expectations became more or less anchored after those breaks. Tables 2– 4 show the regression results
for respectively the euro area, the United Kingdom and the United States, and distinguish three
different sample periods. In each table, the first column refers to the entire sample period, while
columns 2 and 3 refer sub-sample before and after the first break point identified by rolling Chow
tests. Columns 4 and 5 report results for the sub-samples before and after the day on which the Chow
statistic reached its maximum.
The tables highlight that in all three economic areas, inflation expectations became less well anchored
in the run-up to and during the crisis. In particular, the reaction of inflation expectations to news on
euro area inflation was much higher in December 2006–March 2009 than in June 2004–December
17 See also Andrews and Ploberger (1994). Our approach is very similar to that used by Beechey et al. (2008) to test for
parameter instability.
14
2006. The partial R-squared, which excludes the effect of our measures of liquidity and technical
factors, increased from 4.6% to 7.6% between the first and the second sub-sample. The difference is
even starker in the United Kingdom: the partial R-squared increased six-fold, from 1% to 5.7%. Also
US inflation expectations reacted much more to news after the break: 1.8% before March 2006 and
4.9% in the period after.
The change in model fit over time is even more evident when we split the sample period on the dates
on which the Chow test statistics are maximized (columns 4 and 5 in Tables 2–4). The increase in
model performance – measured by the partial R2 – is most visible for the United States, where
economic news explained only 1.7% the of changes in inflation expectations before 14 November
2008 but 26.7% of the variation after that date! The model fit also increases from 2% to 10.8% for the
United Kingdom, and from 4.4% to 16.9% for the euro area. We interpret this result as a warning sign
that central banks should monitor inflation expectations very carefully, and that around the end of
2008, and probably earlier, long-run inflation expectations might have started to be less perfectly
anchored. This is true irrespective of the monetary policy framework and the type of strategy that the
Federal Reserve, the ECB and the Bank of England followed in reaction to the crisis.
To get a sense of the magnitude of the sensitivity of inflation expectations to news, we compared our
results to those reported by Gürkaynak et al. (2006) for the United Kingdom and United States for data
up to 2005. 18 They found that economic news was able to explain 4% of the variation of US inflation
compensation between 1998 and 2005. 19 The results in Table 4 suggest that the degree of anchoring
has declined significantly during the crisis. The same holds for the United Kingdom. Gürkaynak et al.
(2006) report that after the Bank of England became independent, economic news only explained up to
3% of the variation in inflation compensation (between 1998 and 2005). The explanatory power did
not increase when also international news was included. Post October 2008 we find that the degree of
model fit rose to 16.9%, which is almost as high as the model fit that Gürkaynak et al. (2006) found
for the period before the Bank of England gained independence (1993-1997).
6. Conclusions
To what extent have inflation expectations been affected by the economic crisis that erupted in mid2007? In particular, have anchoring properties of long-run inflation expectations changed as the crisis
unfolded in the course of the past two years? In our paper we addressed this question by examining
long-run inflation expectations in the United States, the euro area and the United Kingdom between
June 2004 and March 2009. We considered two types of measures of long-run inflation expectations:
18 Beechey et al. (2007) do not report measures of model fit so we could not compare what we found to their results.
19 This includes the statistically significant effect of several first-business-day-of-the-year dummies, which we did not
include in our regressions.
15
survey-based measures and measures extracted from inflation-linked financial market instruments. The
two measures tell a different story. Survey-based measures remained fairly stable within the central
banks’ comfort zone both during the oil price rally in the run-up to the crisis and as the crisis unfolded.
Expectations measures extracted from inflation-indexed bonds and inflation swaps show a marked
increase in volatility since 2007. Moreover, we found evidence that their sensitivity to news about
inflation and other domestic macroeconomic variables has increased since 2006. The reactivity to
news increased particularly during the period of heightened turmoil triggered by the collapse of
Lehman Brothers in September 2008, when central banks responded by significantly easing monetary
policy using both standard and non-conventional tools.
Although it has been argued that liquidity premia and technical factors have significantly influenced
the behaviour of inflation-indexed markets since the outburst of the crisis, we found that they did not
contaminate the relationship between macroeconomic news and financial market-based inflation
expectations at the daily frequency. We therefore feel confident to draw the conclusion that in all three
economies, long-run inflation expectations have become less firmly anchored during the crisis.
Whether the extent to which anchoring properties have changed during the crisis depends on monetary
frameworks – such as an inflation targeting regime – or the strategies that were followed to counter the
impact of the crisis, is an important topic for future research. In terms of methodology, we conclude
that the approach developed by Gürkaynak (2005) and others is flexible enough to be used to
investigate the properties of expectations during crisis times.
One interpretation of our results is that at the height of the crisis, market participants viewed monetary
authorities as focusing mainly on fixing the monetary transmission mechanism and on softening the
impact of financial instability on the real economy. This appears to have affected market participants’
views on the credibility of central banks in fighting inflation. In the midst of a crisis, this seemed less
of a problem given the absence of inflationary pressures. However, once the economy regains traction,
central banks need to carefully devise exit strategies from the expansionary monetary stance.
16
References
Adam, K. (2007). Experimental evidence on the persistence of output and inflation, The Economic
Journal, 117 (520), 603-636.
Andrews, D.W.K. (1993). Tests for Parameter Instability and Structural Change with Unknown
Change Point. Econometrica, 61 (4), 821-856.
Andrews, D.W.K. and W. Ploberger (1994). Optimal tests when a nuisance parameter is present only
under the alternative, Econometrica 62(6), 1383-1414.
Barclays Capital Research (2008). Global Inflation-Linked Products: A User's Guide.
Bartolini, L., L. Goldberg and A. Sacarny (2008). How Economic News Moves Markets. Federal
Reserve Bank of New York Current Issues in Economics and Finance, 14(6), August.
Beechey, M.J., B.K. Johannsen and A. Levin (2007). Are Long-Run Inflation Expectations More
Firmly Anchored in the Euro Area than the United States?, CEPR Discussion Paper DP6536.
Beechey, M.J. and J.H. Wright (2008). The High-Frequency Impact of News on Long-Term Yields
and Forward Rates: Is It Real?. Federal Reserve Board Finance and Economics Discussion Series,
2008-39.
Bini-Smaghi, L. (2009). Inflation and deflation risks - how to recognise them? How to avoid them?,
Speech at the XXI Villa Mondragone International Economic Seminar, Rome, 24 June.
Brazier, A., R. Harrison, M. King and T. Yates (2008). The Danger of Inflating Expectations of
Macroeconomic Stability: heuristic switching in an overlapping generations monetary model,
International Journal of Central Banking, 4 (2), 219–254.
Clark, T.E. and T. Davig (2008). An Empirical Assessment of the Relationships Among Inflation and
Short- and Long-Term Expectations, Federal Reserve Bank of Kansas City RWP 08-05, November.
Clark, T.E. and T. Nakata (2008). Has the Behavior of Inflation and Long-Term Inflation Expectations
Changed?, Federal Reserve Bank of Kansas City Economic Review, First Quarter.
Deacon, M., A. Derry and D. Mirfendereski (2004). Inflation-indexed Securities: Bonds, Swaps, and
Other Derivatives (2nd edition), Wiley Finance.
Demertzis, M., N. Viegi and M. Marcellino (2007). Inflation targeting – a framework for
communication, DNB Working Paper 149, October.
Demertzis, M., N. Viegi and M. Marcellino (2008). A measure for credibility: tracking US monetary
developments, DNB Working Paper 187, November.
Diebold, F. and C. Li (2006). Forecasting the term structure of government bond yields, Journal of
Econometrics, 130, 337-364.
Diebold, F., G. Rudebusch and B. Aruoba (2006). The macroeconomy and the yield curve: A dynamic
latent factor approach, Journal of Econometrics, 131, 309-338.
Fama, E. and R. Bliss (1987). The information in long-maturity forward rates, American Economic
Review, 77, 680-692.
17
Galati, G., S. Poelhekke and C. Zhou (2008). Did the anchor of inflation expectations in the euro area
turn adrift?, DNB Working Papers 191, December.
Gürkaynak, R., A. Levin, A. Marder and E. Swanson (2006). Inflation Targeting and the Anchoring of
Long-Run Inflation Expectations in the Western Hemisphere, In F. Mishkin and K. Schmidt-Hebbel
(eds.) Series on Central Banking, Analysis and Economic Policies X: Monetary Policy under Inflation
Targeting. Santiago, Chile: Banco Central de Chile.
Gürkaynak, R., A. Levin and E. Swanson (forthcoming). Does Inflation Targeting Anchor Long-Run
Inflation Expectations? Evidence from Long-Term Bond Yields in the U.S., U.K. and Sweden.
Journal of the European Economic Association.
Gürkaynak, R., B. Sack and E. Swanson (2005). The Sensitivity of Long-Term Interest Rates to
Economic News: Evidence and Implications for Macroeconomic Models, American Economic Review,
95(1), 425–436.
Hommes, C., J. Sonnemans, J. Tuinstra and H. van de Velden (2005). Coordination and expectations
in asset pricing experiments, Review of Financial Studies, 18, 955-980.
Hommes, C., J. Sonnemans, J. Tuinstra and H. van de Velden (2007). Learning in Cobweb
experiments, Macroeconomic Dynamics, 11, 8-33.
Jeanne, O. (2009). We need a multilateral consultation on how to avoid global deflation, VoxEU, 3
March.
JP Morgan (2008). Inflation-linked markets, April.
JP Morgan (2009). Global Data Watch, 31 July.
Klaauw, W. van der, W. de Bruin, G. Topa, S. Potter and M.F. Bryan (2008). Rethinking the
measurement of household inflation expectations, FRB of New York Staff Report No. 359.
Levin, A., F. Natalucci and J. Piger (2004). The Macroeconomic Effects of Inflation Targeting,
Federal Reserve Bank of St. Louis Review, 86 (July), 51-80.
Mankiw, N.G., R. Reis and J. Wolfers (2003). Disagreement about Inflation Expectations, NBER
Macroeconomics Annual, 209-248.
Marimon, R. and S. Sunder (1994). Expectations and learning under alternative monetary regimes: an
experimental approach, Economic Theory, 4, 131-162.
Negro, M. del and S. Eusepi (2009). Modeling inflation expectations, Mimeo, Federal Reserve Bank of
New York.
Nelson, C. R. and A.F. Siegel (1987). Parsimonious modeling of yield curves, Journal of Business, 60,
473-489.
Orphanides, A. and J.C. Williams (2005). Inflation scares and forecast-based monetary policy, Review
of Economic Dynamics, 8(2), 498-527.
Paloviita, M. and M. Virén (2005). The role of expectations in the inflation process in the euro area,
Bank of Finland Research Discussion Papers 6.
Pooter, M. de (2007). Examining the Nelson-Siegel Class of Term Structure Models, Tinbergen
Institute Discussion Paper, TI 2007-043/4.
18
Rigobon, R. and B. Sack (2008). Noisy Macroeconomic Announcements, Monetary Policy, and Asset
Prices, In J. Campbell (ed.) Asset Prices and Monetary Policy, NBER conference volume. Chicago:
University of Chicago Press.
Smith, V.L. (2008). Experimental methods in economics, in The New Palgrave Dictionary of
Economics.
Söderlind, P., and L. Svensson (1997). New Techniques to Extract Market Expectations from
Financial Instruments, Journal of Monetary Economics, 40, 383-429.
Stock, J. and M. Watson (2007). Why Has U.S. Inflation Become Harder to Forecast?, Journal of
Money, Credit, and Banking, 39 (February), 3-34.
Svensson, L. (2009). Monetary Policy with a Zero Interest Rate, Speech at SNS, Stockholm, 17
February (http://www.princeton.edu/svensson/papers/MonPolZIR090217e.pdf).
Swanson, E. (2006). Would an Inflation Target Help Anchor U.S. Inflation Expectations?, FRBSF
Economic Letter, Number 2006-20, August.
Zeileis, A., C. Kleiber, W. Krämer and K. Hornik (2003). Testing and dating of structural changes in
practice, Computational Statistics and Data Analysis, 44(1-2), 109-123.
19
Tables
Table 1: Break point dating
Euro Area
Break date:
Chow F-test stat.
p-value
Andrews' test stat.
1% critical value
UK
Break date:
Chow F-test stat.
p-value
Andrews' test stat.
1% critical value
US
Break date:
Chow F-test stat.
p-value
Andrews' test stat.
1% critical value
first date significant
maximum test value
21dec2006
20oct2008
1.950
0.010
4.474
1.59e-09
76.06
33.43
16apr2008
09sep2008
2.302
0.005
5.615
4.93e-10
67.38
26.22
15mar2008
14nov2008
2.236
0.008
13.187
8.99e-26
145.06
24.73
Regressions are run as specified in tables 2 to 4.
20
Table 2: Regression results for Euro Area
(1)
(4)
(5)
(2)
(3)
1st significance
maximum significance
21dec2006
20oct2008
Break date (99% confidence level):
23jun2004 12mar2009
pre-break
post-break
pre-break
post-break
overall R-squared
0.050
0.078
0.091
0.062
0.198
partial R2
0.041
0.046
0.076
0.044
0.169
Observations
1232
651
581
1128
104
France Business Confidence overall indicator
0.018**
0.009
0.030**
0.011*
0.031**
(0.009)
(0.006)
(0.012)
(0.006)
(0.015)
0.022**
0.022***
0.010
0.016**
0.000
(0.009)
(0.008)
(0.013)
(0.007)
(0.000)
France GDP QoQ
France Industrial Production MoM SA 2000=100
France PPI MoM 2000=100
France Unemployment rate SA
France CPI MoM European harmonized NSA
Bundesbank Germany Current Account EUR SA
Germany HICP MoN 2005=100
IFO pan Germany business climate 2000=100
Germany Industrial production MoM SA 2000=100
Germany PPI MoM 1995=100
Germany Unemployment rate SA
Italy Business confidence 2000=100
Italy HICP MoM NSA 2005=100
Italy Industrial Production MoM SA 2000=100
Italy PPI manufacturing MoM 2000=100
Italy Real GDP QoQ SA WDA
∆ Implied Volatility (Euro-bund future continuous call)
∆ VIX (CBOE SPX Volatility (New) – price index)
Constant
0.002
0.000
0.008
0.007
0.005
(0.008)
(0.008)
(0.013)
(0.006)
(0.029)
0.031***
-0.017
0.036***
0.009
0.038***
(0.003)
(0.015)
(0.004)
(0.013)
(0.008)
0.002
0.003
-0.008
0.001
0.000
(0.006)
(0.007)
(0.009)
(0.006)
(0.000)
-0.005
-0.009*
0.009
-0.007
0.128
(0.007)
(0.005)
(0.018)
(0.006)
(0.134)
0.008
0.017**
0.002
0.017***
-0.031
(0.006)
(0.008)
(0.009)
(0.005)
(0.028)
0.009*
0.009**
0.012
0.009*
0.017
(0.005)
(0.005)
(0.018)
(0.005)
(0.015)
-0.007
0.014**
-0.030
0.009*
-0.090***
(0.011)
(0.006)
(0.018)
(0.005)
(0.018)
0.011
-0.003
0.020
0.015
0.013
(0.010)
(0.006)
(0.016)
(0.012)
(0.051)
-0.013
0.009
-0.024
0.008
-0.054**
(0.011)
(0.011)
(0.015)
(0.005)
(0.024)
0.009
0.004
0.015
0.008
0.031
(0.006)
(0.005)
(0.011)
(0.005)
(0.061)
-0.003
-0.001
-0.003
-0.004
-0.013
(0.006)
(0.004)
(0.011)
(0.003)
(0.020)
0.005
0.004
0.005
0.007
-0.027*
(0.005)
(0.006)
(0.008)
(0.005)
(0.016)
0.009
0.013***
-0.001
0.010
-0.103
(0.008)
(0.005)
(0.017)
(0.007)
(0.095)
-0.017*
0.001
-0.033**
-0.006
-0.053
(0.009)
(0.008)
(0.016)
(0.006)
(0.041)
0.032
0.015***
0.037
0.044*
0.044
(0.021)
(0.002)
(0.028)
(0.024)
(0.091)
-0.000
0.008***
-0.004
0.003**
-0.010
(0.002)
(0.002)
(0.002)
(0.001)
(0.006)
-0.005*
-0.006
-0.005
-0.007**
-0.005
(0.003)
(0.004)
(0.003)
(0.003)
(0.005)
0.000
-0.001
0.001
0.000
-0.008
(0.002)
(0.001)
(0.003)
(0.001)
(0.014)
Robust standard errors in parentheses. *** p<0.01, ** p<0.05, *
p<0.1
21
Table 3: Regression results for the United Kingdom
(1)
16apr2008
09sep2008
23jun2004 24mar2009
pre-break
overall R-squared
0.023
partial R2
0.022
Observations
1204
UK industrial production MoM SA
UK CPI EU harmonized MoM NSA
UK retail prices index MoM NSA
UK Nationwide consumer confidence Index SA
UK unemployment claimant count monthly change SA
UK claimant count (unemployment) rate SA
BoE official bank rate
UK chained GDP at market prices QoQ
UK Retail Sales All Retailing
UK PPI Manufactured Products M
UK Avg Earnings Whole Economy
∆ VIX (CBOE SPX Volatility (New) – price index)
Constant
post-break
pre-break
post-break
0.014
0.058
0.028
0.111
0.010
0.057
0.020
0.108
965
239
1066
138
-0.003
0.002
-0.007
-0.001
-0.006
(0.003)
(0.003)
(0.005)
(0.003)
(0.007)
0.007*
0.004*
0.013
0.004*
0.025***
(0.004)
(0.002)
(0.010)
(0.002)
(0.007)
0.009
0.000
0.016
0.004
0.018
(0.016)
(0.006)
(0.031)
(0.006)
(0.031)
-0.000
-0.002
0.013
-0.002
0.025
(0.011)
(0.004)
(0.028)
(0.004)
(0.031)
0.028
no
0.028
-0.047***
0.057***
(0.024)
news
(0.026)
(0.008)
(0.011)
-0.005
-0.004
-0.000
-0.003
0.000
(0.005)
(0.003)
(0.008)
(0.003)
(0.008)
-0.000
0.004
-0.012
0.005
-0.017*
(0.004)
(0.003)
(0.010)
(0.003)
(0.008)
-0.005*
-0.010***
-0.005
-0.010***
-0.005
(0.003)
(0.003)
(0.003)
(0.003)
(0.004)
0.006
-0.003
0.021**
0.002
0.019
(0.006)
(0.005)
(0.009)
(0.006)
(0.015)
0.005
0.000
0.008
0.001
0.020*
(0.004)
(0.004)
(0.007)
(0.004)
(0.011)
0.044**
-0.001
-0.002
0.002
-0.004
(0.003)
(0.001)
(0.008)
(0.003)
(0.019)
-0.001
-0.000
(0.003)
-0.006**
(0.002)
-0.000
(0.003)
-0.007***
(0.002)
(0.002)
∆ Implied Volatility (Euro-bund future continuous call)
(5)
maximum significance
Break date (99% confidence level):
UK Manufacturing PMI Markit survey ticker
(4)
(2)
(3)
1st significance
0.001
0.001
0.000
0.001
0.001
(0.001)
(0.001)
(0.002)
(0.001)
(0.003)
-0.000
-0.001
-0.000
-0.002
0.000
(0.001)
(0.001)
(0.002)
(0.001)
(0.002)
0.000
0.001**
-0.003
0.001**
-0.006
(0.001)
(0.001)
(0.003)
(0.001)
(0.006)
Robust standard errors in parentheses. *** p<0.01, **
p<0.05, * p<0.1
22
Table 4: Regression results for the United States
(1)
Break date (99% confidence level):
15mar2006
14nov2008
pre-break
overall R-squared
0.032
partial R2
0.022
Observations
1189
US Capacity Utilization % of Total Capacity SA
Conference board consumer confidence SA 1985=100
US Industrial Production MoM 2002=100 SA (rate)
US initial jobless claims SA
Conference board US leading index MoM
Federal Funds Target Rate
US new privately owned housing units started by structure total SAAR
(units/thou
US Employees on Nonfarm payrolls total MoM Net Change SA
(thousands)
post-break
pre-break
post-break
0.023
0.063
0.023
0.291
0.018
0.049
0.017
0.267
431
758
1102
87
0.006
-0.009
0.015**
0.005
0.000
(0.006)
(0.012)
(0.006)
(0.006)
(0.000)
0.020*
0.001
0.023
0.010
0.065
(0.011)
(0.012)
(0.016)
(0.010)
(0.049)
0.008
-0.006
0.016
-0.006
0.080***
(0.011)
(0.011)
(0.017)
(0.009)
(0.022)
-0.027*
0.006
-0.033*
-0.023
-0.024
(0.014)
(0.013)
(0.017)
(0.018)
(0.027)
-0.009
0.003
-0.016*
-0.001
-0.033**
(0.006)
(0.004)
(0.008)
(0.003)
(0.016)
0.016
-0.007
0.027
-0.007
0.124
(0.014)
(0.013)
(0.019)
(0.006)
(0.077)
0.009*
0.000
0.010*
0.007
0.041
(0.006)
(0.000)
(0.006)
(0.007)
(0.112)
-0.003
-0.003
-0.004
-0.005
-0.063
(0.005)
(0.006)
(0.010)
(0.005)
(0.275)
0.001
0.011*
-0.007
0.004
-0.025*
(0.005)
Adjusted retail & food services SA total monthly % change
US unemployment rate total in labor force SA
∆ Implied Volatility (Euro-bund future continuous call)
∆ VIX (CBOE SPX Volatility (New) – price index)
Constant
(5)
maximum significance
23jun2004 23mar2009
US Personal Consumption Expenditure Core Price Index MoM SA
(4)
(2)
(3)
1st significance
(0.006)
(0.006)
(0.004)
(0.014)
0.068***
(0.009)
0.014
-0.013*
0.026*
-0.013**
(0.012)
(0.007)
(0.014)
(0.007)
-0.004
-0.015
-0.004
-0.007
0.031
(0.007)
(0.017)
(0.008)
(0.007)
(0.051)
0.001
0.002
0.001
-0.000
0.007
(0.003)
(0.002)
(0.004)
(0.002)
(0.013)
-0.005
0.006
-0.005
-0.003
-0.007
(0.003)
(0.005)
(0.003)
(0.003)
(0.009)
-0.000
-0.001
0.001
0.001
-0.013
(0.001)
(0.002)
(0.002)
(0.001)
(0.015)
Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
23
1
1.5
2
2.5
3
Figure 1: Euro Area Inflation Expectations from Indexed Swaps and the SPF
01jul2004
01jan2006
date
01jul2007
Inflation Expectations, derived
from Inflation Index Swaps
10 year horizon
01jan2009
ECB Survey of
Professional Forecasters
5 year horizon
1
2
3
4
5
Figure 2: Inflation Expectations (vertical line = 15sep2008)
01jul2004
01jan2006
Euro Area
date
01jul2007
UK
01jan2009
US
24
Figure 3: Euro Area Chow Test Value and P-Value
5
1
4
.8
3
.6
2
.4
1
.2
01jul2004
0
.05
.01
01jan2006
21dec2006
date
P-value (left-axis)
01jan2008
01jan2009
Chow F stat. (right-axis)
Figure 4: UK Chow Test Value and P-Value
6
1
4
.8
.6
2
.4
+
.2
01jul2004
0
.05
.01
01jan2006
P-value (left-axis)
date
01jul2007
16apr2008 01jan2009
Chow F stat. (right-axis)
25
Figure 5: US Chow Test Value and P-Value
15
1
14nov2008+
10
.8
.6
5
.4
.2
01jul2004
0
.05
.01
15mar2006
P-value (left-axis)
date
01jul2007
01jan2009
Chow F stat. (right-axis)
26
Appendix 1: The Nelson-Siegel method
There is an extensive literature on modeling the term structure of interest rates or inflation swap rates.
The seminal paper by Nelson and Siegel (1987) proposes a three-factor model that captures the level,
slope and curvature of the yield curve. Empirical studies shows that it fits well for yield curves with
different shapes. 20 At each time point t , the model is given as follows:
(A1)
⎛1− e− λm ⎞
⎛ 1− e− λm
⎞ (m )
− λm
)
y (m
=
β
+
β
+
β
−
e
+ εt .
⎜
⎟
⎜
⎟
t
1,t
2,t
3,t
⎝ λm ⎠
⎝ λm
⎠
)
is the yield with maturity m , (β1,t , β 2,t , β 3,t ) is the vector of parameters for the three factors,
Here y (m
t
which indicate level, slope and curvature of the yield curve, λ is a decay parameter, which usually
)
is an error term.
assumed to be constant across time, and ε(m
t
To estimate this model is by no means an easy task. In particular, taking the dynamics of the betaparameters into account always involves advanced techniques such as Kalman filter (see Diebold et
al.(2006). However, by fixing the decay parameter λ , the estimation becomes a simple OLS
regression. An example of this approach is in Diebold and Li (2006).
We did not choose a particular value for the decay parameter ex ante. Rather, we allowed the decay
parameter to vary on a certain interval, while estimating the Nelson-Siegel model for each specific
value in this interval. We then chose the value of the decay parameter at which the total mean squared
error is minimized.
When estimating the Nelson-Siegel model for a specific value of λ , the observations are the yields at
different maturities. In our data set, the available maturities are 1,2,3,4,5,6,7,8,9,10,12,15,20 and 30
20 For a recent discussion of the empirical performance of the Nelson-Siegel method, see e.g. de Pooter (2007).
27
years. The observations are therefore concentrated at the shorter end of the yield curve, making it more
difficult to accurately represent the shape of the yield curve at longer maturities. In order to balance
this effect, we introduced other maturities – such as 11,13,14,16,17,18,19,21,22 years – . To obtain
yields on those maturities, we used the bootstrapping method in Fama and Bliss (1987).
We followed this procedure to estimate the yield curves for inflation swaps and interest rate swaps for
each day t . Based on this estimated yield curve, we obtained the yield with 9- and 10-year maturities.
These were used to calculate the 10-year ahead forward rates.
28
Appendix 2: Macroeconomic data releases
France
France Business confidence overall indicator
France GDP QoQ
France Industrial production MoM SA 2000=100
France PPI MoM 2000=100
France Unemployment rate SA
France CPI MoM European harmonized NSA
Germany
Germany Current Account EUR SA
Germany HICP MoN 2005=100
IFO pan Germany business climate 2000=100
Germany Industrial production MoM SA 2000=100
Germany PPI MoM 1995=100
Germany Unemployment rate SA
Italy
Business confidence 2000=100
Italy HICP MoM NSA 2005=100
Italy Industrial Production MoM SA 2000=100
Italy PPI manufacturing MoM 2000=100
Italy Real GDP QoQ SA WDA
United Kingdom
Manufacturing PMI Markit survey ticker
Industrial production MoM SA
CPI EU harmonized MoM NSA
Retail prices index MoM NSA
Nationwide consumer confidence Index
Unemployment claimant count MoM SA
Claimant count (unemployment) rate SA
BoE official bank rate
Chained GDP at market prices QoQ
Retail Sales All Retailing
PPI Manufactured Products M
Avg Earnings Whole Economy
United States
Personal Consumption Exp. CPI MoM SA
Capacity Utilization % of Total Capacity SA
Conference board consumer confidence SA
Industrial Production MoM SA (rate)
Initial jobless claims SA
Conference board US leading index MoM
Federal Funds Target Rate
New privately owned housing units started by structure total SAAR
Employees on Nonfarm payrolls MoM SA
Adjusted retail & food services SA MoM
Unemployment rate total in labor force SA
29
Previous DNB Working Papers in 2009
No. 198
No. 199
No 200
No 201
No. 202
No. 203
No. 204
No. 205
No. 206
No. 207
No. 208
No. 209
No. 210
No. 211
No. 212
No. 213
No. 214
No. 215
No. 216
No. 217
No. 218
No. 219
No. 220
No. 221
Peter ter Berg, Unification of the Fréchet and Weibull Distribution
Ronald Heijmans, Simulations in the Dutch interbank payment system: A sensitivity
analysis
Itai Agur, What Institutional Structure for the Lender of Last Resort?
Iman van Lelyveld, Franka Liedorp and Manuel Kampman, An empirical assessment of
reinsurance risk
Kerstin Bernoth and Andreas Pick, Forecasting the fragility of the banking and insurance
sector
Maria Demertzis, The ‘Wisdom of the Crowds’ and Public Policy
Wouter den Haan and Vincent Sterk, The comovement between household loans and real
activity
Gus Garita and Chen Zhou, Can Open Capital Markets Help Avoid Currency Crises?
Frederick van der Ploeg and Steven Poelhekke, The Volatility Curse: Revisiting the
Paradox of Plenty
M. Hashem Pesaran and Adreas Pick, Forecasting Random Walks under Drift Instability
Zsolt Darvas, Monetary Transmission in three Central European Economies: Evidence
from Time-Varying Coefficient Vector Autoregressions
Steven Poelhekke, Human Capital and Employment Growth in German Metropolitan
Areas: New Evidence
Vincent Sterk, Credit Frictions and the Comovement between Durable and Non-durable
Consumption
Jan de Dreu and Jacob Bikker, Pension fund sophistication and investment policy
Jakob de Haan and David-Jan Jansen, The communication policy of the European
Central Bank: An overview of the first decade
Itai Agur, Regulatory Competition and Bank Risk Taking
John Lewis, Fiscal policy in Central and Eastern Europe with real time data: Cyclicality,
inertia and the role of EU accession
Jacob Bikker, An extended gravity model with substitution applied to international trade
Arie Kapteyn and Federica Teppa, Subjective Measures of Risk Aversion, Fixed Costs, and
Portfolio Choice
Mark Mink and Jochen Mierau, Measuring Stock Market Contagion with an Application
to the Sub-prime Crisis
Michael Biggs, Thomas Mayer and Andreas Pick, Credit and economic recovery
Chen Zhou, Dependence structure of risk factors and diversification effects
W.L. Heeringa and A.L. Bovenberg, Stabilizing pay-as-you-go pension schemes in the
face of rising longevity and falling fertility: an application to the Netherlands
Nicole Jonker and Anneke Kosse, The impact of survey design on research outcomes: A
case study of seven pilots measuring cash usage in the Netherlands
DNB Working Paper
No. 35/April 2005
DNB W O R K I N G P A P E R
Jan Kakes and Cees Ullersma
Financial acceleration of booms
and busts
De Nederlandsche Bank